Starting Our Build: A Two-Stage Fact-Checking & Empathy Agent with Google Cloud
As part of the Google Cloud GenAI Academy APAC (Cohort 3) organized by Hack2skill and Google Cloud, we are kicking off a new project to rethink how deep research bots interact with real people.
🎯 The Core Problem
Most current "deep research" systems search dozens of sources, compile the data, and return a dense, 10-page wall of technical text.
When someone is dealing with high-stakes situations—such as navigating government welfare criteria, understanding complex medical trial data, or reviewing tenant legal agreements—an overwhelming dump of jargon causes panic rather than clarity.
We need accurate, cross-verified facts delivered simply and calmly.
🛠️ Planned Tech Stack
- Reasoning & Synthesis: Google Gemini API (via Google AI Studio / Vertex AI)
- Real-Time Fact Retrieval: Native Google Search Grounding with Gemini
- Agent Orchestration: LangGraph / LangChain state graphs
- Data & History: Supabase
🚀 Next Steps
We are currently building the state-graph flow in LangGraph to handle the search-and-verify loops before piping the data into the synthesis layer. Updates and prototype demos will follow as development progresses.
Top comments (0)